System Detail

wespeaker/ res34-voxceleb

wespeaker_resnet34_voxceleb

ranked review: approved WeSpeaker ResNet

This page shows the global rank, scenario ranks, and trial-set results for this submission.

Back To Leaderboard Compare Model Variants Submit Another Run

Global Position

#7/18

Public ranking position among reviewed full-core submissions.

Scenario-Macro EER

12.0516

Scenario-Macro minDCF

0.477054

Coverage

26/26

0 scenarios ranked #1 0 scenarios in top 3

Public Listing

Ranked Publicly

This full-core result is approved and included in the public ranking.

Model Provenance

Source and architecture

Displayed name: wespeaker/res34-voxceleb
Source: WeSpeaker
Architecture: ResNet
Submitted alias: wespeaker-voxceleb-resnet34-LM
Created at: 2026-05-29T14:51:34.068325+00:00

Training And Links

Training data and references

Training data: VoxCeleb2 dev, 5,994 speakers
Training setup: ResNet34-TSTP-emb256 r-vector, large-margin fine-tuned; official card reports 6.63M parameters and 4.55G FLOPs.
Submission mode: full-core
Paper: Deep Residual Learning for Image Recognition

Open Source Checkpoint Open Paper

Higher-Ranked Scenarios

Scenario ranks above the global position

Aging #6 Speaking style #6 In-the-wild #6

Show 3 supporting trial sets

GLOBE: #4 on its trial-set ranking.
Lombard Grid Lombard: #5 on its trial-set ranking.
ESD: #6 on its trial-set ranking.

Lower-Ranked Scenarios

Scenario ranks below the global position

Channel/device #10 Short-duration #9 Distance #9

Show 3 supporting trial sets

CHiME-6 Domestic Far-Field: #10 on its trial-set ranking.
GSC Short: #10 on its trial-set ranking.
AliMeeting Overlap: #10 on its trial-set ranking.

Scenario Rankings

Full ranking across scenarios

This table shows where the model sits on each scenario, using source-balanced scenario scores and the linked trial sets as evidence.

Scenario	Rank	Lens	Evidence	Score
Aging Robustness How stable identity representations remain across age-derived and longitudinal recording gaps.	#6/18 Competitive	Speaker aging and time gaps 1.4319 EER from leader	#6 4.9242 #7 2.0788	3.5015 0.146062 minDCF
Speaking-Style Robustness Whether the model can preserve identity across emotion-driven change, whispered speech, and noise-induced Lombard speaking style.	#6/18 Competitive	Speaking style shift 2.2464 EER from leader	#6 3.5211 #9 11.4375 #5 0.1490	5.0359 0.303656 minDCF
In-The-Wild Robustness How strong the model is on unconstrained celebrity and media speech across official CN-Celeb and VoxCeleb protocols.	#6/18 Competitive	Open-domain media speech 1.3742 EER from leader	#7 11.9718 #8 0.8188 #8 0.9347 #7 1.6768	6.5576 0.240625 minDCF
Overlap Robustness How well speaker identity survives light, mid, and heavy overlap in both meeting and domestic recordings.	#7/18 Competitive	Overlapping speakers 3.8031 EER from leader	#10 9.9940 #9 17.8889	13.9414 0.591408 minDCF
Cross-Lingual Robustness How well speaker identity survives enrollment-test language mismatch.	#8/18 Competitive	Language mismatch 1.0856 EER from leader	#8 5.5532	5.5532 0.361444 minDCF
Noise / Reverb Robustness How reliably the model preserves identity when room acoustics, distractor noise, and microphone placement deviate from an easier in-corpus reference condition.	#8/18 Competitive	Noise and reverberation 3.3480 EER from leader	#8 5.9280	5.9280 0.181180 minDCF
Accent / Dialect Robustness How reliably the model tracks identity across accent and dialect mismatch.	#8/18 Competitive	Accent and dialect variation 1.9922 EER from leader	#4 25.0600 #8 19.7947	22.4273 0.700394 minDCF
Genre-Shift Robustness Whether performance holds when CN-Celeb enrollment and test speech come from different source genres.	#8/18 Competitive	Source-genre variation 12.2400 EER from leader	#8 26.5125	26.5125 0.822805 minDCF
Distance Robustness How much performance changes across meeting and domestic distance mismatch.	#9/18 Competitive	Distance mismatch 3.7329 EER from leader	#9 18.6760 #9 8.8800 #10 13.6871 #8 10.7439	12.9967 0.561170 minDCF
Short-Duration Robustness How much performance holds up when speech evidence is limited by duration.	#9/18 Competitive	Short-duration speech 2.3936 EER from leader	#9 7.0523 #10 13.2190 #7 25.4223 #7 13.9382	14.9080 0.610039 minDCF
Channel / Device Robustness How resilient the model is to device and channel mismatch.	#10/18 Needs work	Device and channel variation 8.8947 EER from leader	#9 19.8267 #8 10.5833	15.2050 0.728816 minDCF

Variant Compare

Compare ResNet variants

Expand this section to compare checkpoints in the same architecture group by source, training data, training setup, and leaderboard result.

Show

Variant	Source	Training Data	Training Setup	Global Rank	Macro EER	Macro minDCF	Open
wespeaker/res152-voxceleb	WeSpeaker	VoxCeleb2 dev, 5,994 speakers	ResNet152-TSTP-emb256 r-vector, ArcMargin, 150-epoch VoxCeleb2 recipe with speed perturbation and MUSAN/RIRS augmentation.	#4 ranked	11.2275	0.442748	Open
wespeaker/res293-voxceleb	WeSpeaker	VoxCeleb2 dev, 5,994 speakers	ResNet293-TSTP-emb256 r-vector, large-margin fine-tuned; official card reports 28.62M parameters and 28.10G FLOPs.	#5 ranked	11.3130	0.439607	Open
wespeaker/res34-voxceleb Current	WeSpeaker	VoxCeleb2 dev, 5,994 speakers	ResNet34-TSTP-emb256 r-vector, large-margin fine-tuned; official card reports 6.63M parameters and 4.55G FLOPs.	#7 ranked	12.0516	0.477054	Open
wespeaker/res34-cnceleb	WeSpeaker	CN-Celeb train	ResNet34 r-vector with TSTP pooling and large-margin fine-tuning on the CN-Celeb WeSpeaker recipe.	#14 ranked	14.4910	0.607964	Open

Trial-Set Results

Raw ranking by trial set

Expand this section to inspect the detailed trial-set rankings.

Show

Trial Set	Rank	Scenarios	Trials	Score
TidyVoiceX2-ASV TidyVoiceX2-ASV	#8/18	Cross-lingual	200000	5.5532 0.361444 minDCF
HI-MIA HI-MIA	#9/18	Short-duration	660000	7.0523 0.323703 minDCF
GSC Short Google Speech Commands	#10/18	Short-duration	220000	13.2190 0.705995 minDCF
GLOBE GLOBE	#4/18	Accent/dialect	56848	25.0600 0.531734 minDCF
CN-Celeb CN-Celeb	#7/18	In-the-wild	3484292	11.9718 0.411595 minDCF
CN-Celeb Genre CN-Celeb	#8/18	Genre shift	440000	26.5125 0.822805 minDCF
CN-Celeb Short CN-Celeb	#7/18	Short-duration	546964	25.4223 0.914387 minDCF
VoxCeleb1-O VoxCeleb1	#8/18	In-the-wild	37611	0.8188 0.052968 minDCF
VoxCeleb1-E VoxCeleb1	#8/18	In-the-wild	579818	0.9347 0.057497 minDCF
VoxCeleb1-H VoxCeleb1	#7/18	In-the-wild	550894	1.6768 0.098504 minDCF
VoxCeleb Short VoxCeleb1	#7/18	Short-duration	394724	13.9382 0.496071 minDCF
3D-Speaker Device 3D-Speaker	#9/18	Channel/device	180000	19.8267 0.860520 minDCF
3D-Speaker Distance 3D-Speaker	#9/18	Distance	175163	18.6760 0.836862 minDCF
3D-Speaker Dialect 3D-Speaker	#8/18	Accent/dialect	180000	19.7947 0.869053 minDCF
FFSVC 2022 Cross-Channel FFSVC 2022	#8/18	Channel/device	72000	10.5833 0.597111 minDCF
FFSVC 2022 Cross-Domain FFSVC 2022	#8/18	Distance	66546	10.7439 0.612159 minDCF
Whisper40 Whisper Whisper40	#9/18	Speaking style	17600	11.4375 0.645688 minDCF
Lombard Grid Lombard Lombard Grid	#5/18	Speaking style	29524	0.1490 0.010991 minDCF
VOiCES Noise/Reverb VOiCES	#8/18	Noise/reverb	55000	5.9280 0.181180 minDCF
CHiME-6 Domestic Far-Field CHiME-6	#10/18	Distance	36487	13.6871 0.500573 minDCF
CHiME-6 Overlap CHiME-6	#9/18	Overlap	39600	17.8889 0.651583 minDCF
ESD ESD	#6/18	Speaking style	437408	3.5211 0.254288 minDCF
AliMeeting Near/Far AliMeeting	#9/18	Distance	220000	8.8800 0.295085 minDCF
AliMeeting Overlap AliMeeting	#10/18	Overlap	165000	9.9940 0.531233 minDCF
VoxKnesset VoxKnesset	#6/18	Aging	158312	4.9242 0.199224 minDCF
VoxPopuli Aging VoxPopuli	#7/18	Aging	146575	2.0788 0.092901 minDCF